Roots of Gender Equality: the Persistent Effect of Beguinages on Attitudes Toward Women

A. Frigo and E. Roca Fernandez

Discussion Paper 2019-13

Roots of Gender Equality:

the Persistent Effect of Beguinages on Attitudes Toward Women

Annalisa Frigo1 and Eric Roca Fernandez∗2

1IRES/LIDAM, UCLouvain

2Aix-Marseille Univ., CNRS, EHESS, Centrale Marseille, AMSE, Marseille, France

Abstract

This paper is concerned with the historical roots of gender equality. It proposes and

empirically assesses a new determinant of gender equality: gender-specific outside options

in the marriage market. In particular, enlarging women’s options besides marriage —even

if only temporarily— increases their bargaining power with respect to men, leading to a

persistent improvement in gender equality. We illustrate this mechanism focusing on Belgium,

and relate gender-equality levels in the 19th century to the presence of medieval, female-only

communities called beguinages that allowed women to remain single amidst a society that

traditionally advocated marriage. Combining geo-referenced data on beguinal communities

with 19th-century census data, we document that the presence of beguinages was instrumental

in decreasing the gender gap in literacy. The reduction is sizeable, amounting to a 5.3% drop

in gender educational inequality.

Keywords: Economic Persistence, Culture, Institutions, Religion, Gender Gap.

JEL: I25, J16, N33, O15, O43, Z12.

∗We thank David de la Croix, Matteo Cervellati, Oded Galor, Marc Goni, Fabio Mariani, Felipe ValenciaCaicedo, and all the participants to seminars at Brown University, Universite catholique de Louvain, Universitedu Luxembourg, the Indian Statistical Institute Delhi, Syddansk Universitet and Universita Commerciale LuigiBocconi for precious comments and suggestions. Annalisa Frigo acknowledges the financial support from theBelgian French-speaking Community (convention ARC no15/19-063 on “Family Transformations: Incentives andNorms”). This work was also supported by French National Research Agency Grant ANR-17-EURE-0020.

1

1 Introduction

The role of women in the society is heavily influenced by cultural norms, which prescribe —among

others— whether they can work, earn income and, in general, enjoy the same rights as men’s.

Scholars have shown that gender equality positively impacts economic outcomes ranging from

historical human capital accumulation (Carmichael et al. (2016)) to present-day economic growth

(Klasen (2002) and Klasen and Lamanna (2009)). Moreover, attitudes toward women seem to be

very persistent (Fernandez and Fogli (2009)) and their origins have been traced back to cultural

shocks and gender-specific physiologic endowments —see, among others, Becker and Woessmann

(2008), Grosjean and Khattar (2017), and Alesina et al. (2013).

This paper is concerned with the deep determinants of gender equality and puts forward

a new mechanism to explain variations in social norms regarding the role of women in society.

In particular, we propose a new economic explanation related to female bargaining power in

the marriage market and how alternative opportunities for women increase it. We illustrate the

relevance of such mechanism focusing on the Belgian context, where the historical presence of

medieval, female-only, semi-religious communities called beguinages allowed women to remain

single and independent of men in stark contrast to the customs of the time. Beguines also

enjoyed financial freedom since they supported themselves working for pay and, contrary to

nuns, retained and accumulated wealth. Combining information on the geographic location of

beguinages with Belgian census data, we show that municipalities where beguine communities

thrived reached higher levels of gender equality by the mid-19th century. In particular, we

find a 5.3% reduction in the gender literacy gap, our preferred measure of gender equality. We

address the possibility that beguines’ location decisions obeyed pre-existing local conditions in

terms of gender-egalitarian attitudes relying on a series of robustness tests: alternative variable

coding, the adoption of an exogenous variable that leverages towns’ network, an IV approach,

and propensity score matching.

This study supports the idea that gender roles and gender equality are strikingly persistent

and determine present-day attitudes towards women.1 However, our explanation departs

from those advanced in the existing literature, which typically ascribe the origins of gender

roles to gender-based specialisation predating industrialisation. For instance, Alesina et al.

1In general, cultural beliefs are deeply rooted and tend to evolve slowly. Several studies document persistencefor a wide range of characteristics, including antisemitism and trust. See, for instance, Voigtlander and Voth(2012), Becker et al. (2016) or Lowes et al. (2017). Giuliano and Nunn (2017) analyse under which conditionsculture persists or changes.

2

(2013) maintain that gender-specific endowments facilitated economic specialisation, prompting

gender roles. The comparative advantage of men in manoeuvring the plough explains why the

descendants of plough-agriculturalists are today less gender-egalitarian. Similarly, Hansen et al.

(2015) show that today women work less in societies that transitioned earlier to agriculture.

The proposed channel operates through differential fertility across modes of living, diverting

agriculturalist women’s time from market work to childcare. Galor and Weil (1996) propose

a theoretical framework encompasing the previous findings. It emphasises the importance of

differential physical endowments —in terms of strength— and their complementarity to physical

capital. The endowment imbalance between genders generates economic specialisation as long

as physical capital is absent: men work and women bear children.

Besides physiological determinants, scholars have shown that certain cultural and economic

aspects are associated with a gender bias. Alesina and Giuliano (2010) relate strong family ties

to gender inequality, although probably both variables co-evolved.2 Grosjean and Khattar (2017)

stress that historical scarcity of women improved their position in Australia during the 17th

and 18th century. Today, these traditional roles persist.3 Fernandez and Fogli (2009) analyse

the behaviour of second-generation migrants in the US: female labour force participation in the

parents’ country of origin explains the behaviour of their daughters. Finally, Giuliano (2018)

provides an extensive literature review on the historical origins of gender roles.

The two-step mechanism we postulate, instead, combines female empowerment through

girls’ bargaining power —facilitated by the presence of beguinages— with the intergenerational

transmission of a gender-egalitarian culture. Initially, the very existence of beguinages provides

an additional option to girls: besides marriage or monastic life, they were given the possibility

of remaining single and financially independent. The availability of a viable outside option (to

marriage) improves the bargaining power of women, which in turn brings about a progressive

reduction of the gender gap. The improved status of women favours the emergence of more

gender-egalitarian norms, which are then perpetuated across generations through a cultural

transmission mechanism a la Bisin and Verdier (2001). In that sense, children in towns where

beguines located were exposed firstly to beguines themselves, and secondly to peers who were

already more favourable towards gender-equal norms. Role modelling is a possible competing

2Bertocchi and Bozzano (2019) validate this result using historical Italian data. Tur-Prats (2018) associatesthe stem family type to domestic, intimate-partner violence in Spain.

3Their argument builds on the relative shortage of women, while this is irrelevant in our setting. However, weboth emphasise the importance of the marriage market.

3

channel. We can partly dismiss it: we find that the sheer accessibility to beguinages explains

gender equality. Hence, the transmission of gender norms operates even absent direct contact with

beguines. Instead, viable alternatives to marriage is pivotal in delivering gender equality: non-

enclosed feminine monasteries also enhance women’s position. These offer a more constrained

alternative to marriage —compared to beguinages— but definitely lack the role modelling

dimension.

By focusing on semi-monastic communities, our paper is also linked to the strand of the

literature analysing the legacy of religious institutions on economic performance. For instance,

Valencia Caicedo (2018) assesses the positive effects of historical Jesuit Missions in Latin America

on modern-day human capital and technological progress. Waldinger (2017) related the presence

of Catholic missions in Mexico to increased literacy and educational attainment. Similar results

extend also to Protestant missionaries in Africa, as in Cage and Rueda (2016). Andersen

et al. (2017) associate the presence of the industrious Cistercian Order in England with faster

productivity growth long after monasteries were disbanded. Finally, Akcomak et al. (2016)

study how a Catholic sect that valued literacy to understand the Gospel —the Brethren of the

Common Life— affected Dutch male literacy rates.

Our study is also related to the literature focusing on intra-couple bargaining. Female

education reinforces their status and aligns family preferences to women’s, as described by

Iyigun and Walsh (2007) and Doepke and Tertilt (2009). Giuliano (2018) argues that matrilineal

societies favour gender equality by providing a safety net for women, raising women’s bargaining

power. Gottlieb and Robinson (2016) and Jayachandran (2015) discuss this hypothesis. Our

paper bridges the gap between these two strands of the literature, suggesting that even a

temporary improvement in women’s bargaining position can have long-lasting effects on gender

equality.

To the best of our knowledge, this paper is the first to systematically explore the relationship

between pre-industrial, female-only urban associations and gender equality in the long run. Our

contribution singles out the importance of female associations in improving gender equality,

well beyond the life-cycle of the original institutions. Moreover, we propose a novel mechanism

based on women’s bargaining power, never explored in this literature, to explain the enduring

association between female communities and gender equality.

The rest of the paper is organised as follows: Section 2 introduces the historical background

of the beguinal movement. Sections 3 and 4 describe the different data sources and the empirical

4

strategy. The latter is supplemented by robustness tests, including an IV approach. A theoretical

model in Section 5 illustrates the suggested mechanism that rationalises our findings. Section 6

concludes.

2 Historical Background

The beguinal movement started in the late 12th century in the Northern Sea region,4 fully

flourishing in the next two centuries, and has left several beguinages across the Low Countries,

France and Germany.5 It consisted of self-supporting, female-only, semi-religious communities

that offered single women of all social strata the opportunity to work and conduct an independent

life in a society that disapproved that option.6

Records of pre-beguine communities first appear around the city of Huy around 1181. Several

analogous groups arose nearby in subsequent years, dotting the region comprising Aachen,

Nivelles and Vilvoorde. Simultaneously, two beguinages flourished in Cambrai and Valenciennes

—in present-day France—, see Simons (2010, Appendix I). Bowie’s (1990, p. 14) chronology is

similar and also points to the same Belgian region as the cradle of the beguine movement.7 From

its early configuration as groups of itinerant, devote women, the movement settled in definitive

spaces within towns. We can distinguish two main types of beguinages: convent and court.8 In

the first sort, the most common, beguines lived close to each other in the same neighbourhood

4The economic literature argues that the gender-egalitarian culture of the Northern Sea region, documentedin historically high levels of female employment (De Moor and Van Zanden (2010)), was conducive to economicdevelopment. In particular, gender equality enabled human capital accumulation by aligning family’s fertilitywith women’s preferences for less offspring (Carmichael et al. (2016)). However, we do not attempt to explainwhy that region was more gender-egalitarian, to begin with. Instead, we are concerned with differences in genderequality within this area.

5In 1998, the UNESCO recognised some of them as World Heritage for their “outstanding example [... as] areligious movement characteristic of the middle ages, associating both secular and conventual values”.

6Poor and well-off women joined beguinages alike. Destitute girls and women who sought poverty as a Catholicideal became members, see McDonnell (1954, p. 146) and Geybels (2004). On the other hand, daughters ofwealthy citizens and noble-men entered the community, at times accompanied by their maids, see McDonnell(1954, p. 97-99) and Geybels (2004, p. 111).The Mistresses of the beguinages in Ghent indicate that “young beguines received their education [...]” in thebeguinage, although the “fruit of their work” earned them a living, see Vander Schelden (1862, p. 2 and p.12); own translation from French. Beguinages combined contemplative lifestyle and work (McDonnell (1954, p.146-149)). Indeed, admittance to a beguinage required the capacity of supporting oneself through work or personalfinances (Geybels (2004, p. 114-118)) and beguines’ laboriousness put them “in competition with guilds andcraftsmen” (Ibid. p. 63). Beguine occupations included “carding, spinning, weaving and bleaching [... ,] makingof bobbin lace, sewing, embroidery, making rosaries and baking communion wafers”. “[L]abour was compatiblewith the ideals of the vita apostolica [... because] the disciples in Jesus’s day also had to support themselves”.(Ibid. p. 115-116). Caretaking and teaching were other profitable sources of income.

7These initial arrangements differed from the concept of beguinage that developed later. However, theyfeatured several women who followed a spiritual life outside convents.

8Geybels (2004, p. 105-106) distinguishes between “town pattern” beguinages, “court beguinages” and whathe calls “miscellaneous beguinages”.

5

and followed a common rule under the direction of a superior.9 Although court beguinages

were rarer, this type developed earlier and left a greater imprint on city life. In fact, court

beguinages were erected with the purpose of accommodating the community and consisted of

several dwellings, gardens, workshops and a chapel.

Figure 1: Number of beguinages created by decade.

1250 1300 1350 1400 1450 1500 15500

5

10

15

20

25

30

35

Year

Beg

uina

ges Total beguinages

Convent beguinages

Court beguinages

Note: Data on beguinage foundations per decade from Simons (2010), Table 1. Own elaboration.

The expansion of beguinages during the middle ages was uneven. Across the Low Countries,

after an initial surge, the creation of beguinages stagnated at 20 per decade and then steadily

decreased. Figure 1 represents the number of beguinages created every decade between 1241

and 1565. Overall, 111 cities hosted a beguinage: 70 located in Belgium, 22 in France, 17 in the

Netherlands, and 2 in Germany. The number of municipalities that had at least one beguinage

over time is depicted in Figure 2 separately for Belgium and for all countries. After an initial

momentum in the creation of new beguinages, the movement reached its zenith between the 14th

and the 16th centuries, when the number of beguinages stabilised. Afterwards, their popularity

declined.

Beguinages differed greatly in terms of the population they hosted. Estimates are scarce

but attest to the popularity of the movement.10 For instance, Mechelen’s beguinage hosted up

to 1900 beguines in 1500 while that of Nivelles counted 51 beguines in 1284, shortly after its

creation. However, historians document a decreasing trend between 1300 and 1900. By the

mid-19th century those beguinages still operating counted less than 10 inhabitants.11

9See, among others, Geybels (2004, p. 50) and Mommaers and Dutton (2004, p. 21).10See Simons (2010, Table 3, p. 56-59; Appendix II), McDonnell (1954, p. 64) and Geybels (2004, pp. 69, 136,

143, 145 and 147).11Some pious men, known as beghards, emulated the lifestyle of beguines but never reached their reputation

and size. Beghards were sorts of small guilds, as they were essentially craftsmen in the textile industry.

6

Figure 2: Number of cities with at least one beguinage.

1300 1400 1500 1600 1700 1800Year

20

40

60

80

Cities with beguinage

All countries

Only Belgium

Note: Data on beguinage foundations per decade from Simons (2010), Table 1. Own elaboration.

Although beguinages were preferentially established in relatively larger towns, communities

also settled in hamlets or small boroughs counting less than 200 houses in 1437. This was the

case of Bousval, Maleves-Sainte-Marie-Wastinnes, Thorembais-les-Beguines and Grez-Doiceau.

At that time, the average town had 187 dwellings,12 and the larger —Brussels, Leuven and

Antwerp— counted 6376, 3579 and 3440 houses.

The beguine movement stands out from other religious communities especially because

beguines never constituted an Orthodox religious order.13 Albeit religiosity and pious life

guided the beguinages’ foundation and daily routine,14 these communities radically differed from

monastic convents. First, beguines did not take poverty vows, allowing them to retain property

and accumulate wealth. Second, they could leave the beguinage at will to marry or to reintegrate

the secular society. Family ties were generally maintained.15

Third, these communities were actively involved and well integrated into the everyday social

life of medieval cities. As thoroughly depicted by Simons (2010), whether court beguinages were

intended to do so or not, they attracted women from rural areas who could find work there in a

safe and supportive environment. Alongside prosperous beguines lived others of more modest

12The data on municipal size during 1437 is from Cuvelier (1912) and is restricted to the historical Brabantprovince. It comprises 439 towns, of which 22 had a beguinage.

13To Denisart, “beguines are pious women whose status is [...] between religious and secular”; own translationfrom French, see Vander Schelden (1862, Document V, p. XXV-XXVI).

14McDonnell (1954, p. 5) and Bethune (1883, p. 90).15Warnkonig (1836, p. 350) defines them as “women [...] who wear a particular monastic clothing, live

together in a space not forbidden to outsiders [...], to lead a chaste and pious life under a superior and followingtheir own statutes. They are not obliged by vows to renounce the world forever”; own translation from French.Vander Schelden (1862, Document V, p. XXV-XXVI) confirms this characterisation adding that “beguines neednot give up family relationships or friendships [...]”. Bishop Malderus of Antwerp similarly characterises beguines,especially noting the lack of vows, see McDonnell (1954, p. 121-122). Mommaers and Dutton (2004, p. 20) portraya similar arrangement.

7

background who supported themselves as teachers, maids, nurses, textile workers, or small

business owners. In fact, while beguines promised to lead a humble life, this did not prevent

them from having real-world occupations.16 Beguines served in hospitals and leper houses, and

the nursing of individuals in private homes, both for charitable reasons and in order to earn a

small income. Some beguines were farmers in nearby fields, herded animals or grew vegetables

for the local market. The bulk of the evidence, however, suggests that most beguine workers

were labourers in the textile industry and earned their living from commerce: for instance, a

few traded cloth locally and interregionally in Bruges, Ghent, and Arras, while in Tournai will

records provide documentary evidence of a draper employing forty fellow beguines.17

The movement further distinguished from Catholic orders because beguines lacked common

rules and a central coordinating authority.18 Despite their non-adherence to traditional monastic

standards and although communities were independent of any male authority, beguines were

tolerated and even granted spaces by the secular society.19 In general, the Holy See and local

bishops supported the beguine movement.20 However, the Council of Vienna (1311-12) demonised

beguines, and the persecution of heretic movements harassed them particularly in the Rhineland

region, despite the fact that the beguine movement was never considered as such.21 Accusations

did not last long, though, and the Pope Clement VI completely exonerated the movement from

previous charges in 1328.

Beguines’ peculiar freedom of choice and self-determination rapidly propagated and, as

soon as 1260, a literary work portrayed the beguines among other religious orders.22 Indeed,

beguinages were the forerunners of women’s associations, guaranteeing the necessary social and

16McDonnell (1954, p. 131) provides an analogous characterisation based on Bethune (1883, p. 74): “[t]hebeguinage was a retreat, especially well adapted to an urban society, where women living in common could pursuechastity without a vow and earn a livelihood by suitable work”. See also Bowie (1990, p. 25).

17McDonnell (1954, p. 5) underlines their charitable work.18According to Geybels (2004, p. 59), the beguine movement was “poorly organised [...] with each community

possessing its own statutes”. It always lacked common rules (McDonnell (1954, p. 134)) and ecclesiasticalopposition was partly founded on this, see Fredericq (1889, p. 168-169, par. 172).

19French and Belgian beguines enjoyed powerful patronage and were granted privileges, including land donations,see Vander Schelden (1862) and McDonnell (1954, p. 6, Ch. X and Ch. XI). Clerical tolerance is also emphasized,and beguines’ lifestyle was praised and recommended by prelates and the Roman Curia.

20Simons (2010, p. 124-125) indicates that clergymen accepted beguines as “a sign of diversity” within theChurch, and others simply saw beguines as women who “contributed to purify the faith”. However, some initialopposition is recorded because they preached and “interpreted the mysteries of Scripture”. See also footnote 19.

21See Fredericq (1889, p. 168-169, par. 172) and Geybels (2004, p. 56-65).22The poem “Ci encoumence li diz des Beguines” written by the Parisian trouvere Rutebeuf around 1260

caricatures beguines “If a beguine marries, that is her vocation, because her vows or profession are not for life.Last year she wept, now she prays, next year she’ll take a husband. Now she is Martha, then she is Mary; nowshe is chaste, then she gets a husband.”. The author of this piece and other foreign visitors in the southern LowCountries expressed surprise over the public behaviour of women: they were usually shocked by women’s relativefreedom of movement and action.

8

financial protection for single women who joined the community, in a society where staying

single was stigmatised.23 Because of their independence from male authority, Bowie (1990, p.

13) considered the beguines “the first European women’s movement”.

Finally, De Moor (2014) reviews and dismisses the traditional explanations for the rise and

success of the beguine movement. More in detail, the view that beguines sheltered urban girls

unable to marry because of skewed urban sex rations is dubious because most beguines came

from the countryside. Similarly, beguinages were not a substitute for monasteries, because each

community located geographically separated and attracted different women. The author evinces

that —besides devotional reasons— beguines were often motivated by professional opportunities.

In that sense, joining a beguinage was a deliberate choice to stay single and do remunerated

work, independently of a male breadwinner.

3 Data

We combine several data sets to establish an empirical relationship between medieval beguine

communities and female agency while accounting for a rich set of potential confounders.

Historical Presence of Beguines. Beguinage location, foundation and dissolution dates

are reported by Simons (2010). The author combines a wide variety of sources to overcome

the lack of historical records and the fact that beguinages were rarely established through a

formal act of foundation. He documents 143 sites that can be positively defined as beguinages,

with few records grouping multiple small beguinages in the same town. Using this information,

we compile a new data set that features the exact location of each beguinage and relate it to

Belgian municipalities existing in 1866. We restrict our attention to beguinages located within

the borders of modern Belgium, where 70 municipalities hosted at least a community. Figure 3

depicts their location.

Ideally, one would like to have information on the number of beguines living in each

municipality, but such data is scarce and only available for selected beguinages. We circumvent

this limitation by creating two proxy variables that convey the intensive and extensive margin

of such presence. The first one is a binary indicator that takes value one when a municipality

23The cartulary of the beguinage in Ghent supports this perspective, see Geybels (2004, p. 47).

9

Figure 3: Distribution of beguinages and literacy equality index.

Literacy Equality Index (1866) 0.3167 - 0.4180 0.4180 - 0.4337 0.4337 - 0.4409 0.4409 - 0.4475 0.4475 - 0.4554 0.4554 - 0.4625 0.4625 - 0.4678 0.4678 - 0.4743 0.4743 - 0.4858 0.4858 - 0.6439 BeguinageCommunes given to Belgium following the Treaty of Versailles

Legend

0 50 100 km

Note: This map depicts present-day Belgian municipalities. The literacy equality index is computed for eachhistorical municipality during 1866 and aggregated using population-weighted averages. Provincial boundariesfollow historical limits. Due to the skewness of the data, the map displays it by deciles.

had, at any point in time and for any duration, at least one beguinage:

Beguinagei =

0 if municipalityi never had a beguinage,

1 if municipalityi had, at least, one beguinage.

(1)

The second measure consists of a continuous variable that accounts for different durations (t) of

beguinages (j) in municipalities (i). Since it is possible that a municipality hosted more than one

beguinage, we aggregate the impact of different beguine communities coexisting or succeeding in

the same town in the variable exposure:

Exposurei =∑j∈i

ti,j . (2)

Census Data. Following Bertocchi and Bozzano (2019), we use census data to assess gender

equality comparing female-to-male literacy rates.24 This source offers widespread coverage and

consistent measurement across regions, with a final sample comprising 2553 municipalities with

complete information. Data to proxy gender equality with literacy ratios are only available

at the municipal level starting from the 1866 census.25 Analysing 19th-century census data

mitigates the influence of the labour market and welfare state measures implemented in the

following century, especially compulsory education. Moreover, it allows to control for migratory

patterns and reduces concerns related to the involvement of women in the early industrialisation

24In our context, a literate individual is someone able to read and write.25Census data from 1846 onwards have been recently digitised by the HISSTAT/LOKSTAT project of the

University of Ghent. Older national censuses from the years 1800-1830 are archived as manuscripts in Belgium,France and the Netherlands.

10

of Belgium. Finally, historical data alleviates the fact that variation in gender norms tends to

be confounded over time by other events and institutional shocks.

In the first half of the 19th century, education was not compulsory for children and traditional

roles preferentially educated sons over daughters. Therefore, comparing female-to-male literacy

rates is meaningful as it reflects how women were perceived with respect to men at a time in

which schooling was an expensive investment.26 Locations valuing women comparatively more

and equating them to men are expected to educate both genders in similar manners.

We proxy gender equality by taking the ratio of literacy levels between women and men. For

each municipality i, it amounts to:

literacy equality indexi =Number of literatewomeniNumber of literatemeni

(3)

which relates the number of literate women to the number of literate men.

Alternative definitions of this ratio are considered for sensitivity checks. These are, for each

municipality i :

female literacy sharei =Number of literatewomeni

Number of literatewomeni +Number of literatemeni, (4)

recording the share of literate women within the literate population, and

female literacy indexi =Share of literatewomeniShare of literatemeni

(5)

comparing female and male literacy rates. Higher values of any index correspond to more

favourable attitudes towards women. Notice that the inclusion of men in the denominator is

meaningful since plain literacy rates may vary across regions for diverse reasons. Its inclusion

smooths out part of these underlying, idiosyncratic characteristics.

Other variables derived from the census comprise population, nuptiality and the share of

international and domestic migrants by gender, the sex ratio and an indicator variable for the

10% most populous municipalities.

Education Providers. The empirical analysis takes into account the presence of two

education providers that promoted literacy and hence could confound our outcome of interest.

First, elementary education was traditionally offered almost exclusively by church-related

institutions prior to and during the 19th century. According to Ruyskensvelde (2016, p. 28),

schooling became compulsory at the end of the 19th century, whereas previous efforts —notably

the Nothomb law of 1842— failed.27 Our data set documents the presence of monasteries at

26Clearly, the parental decision on whether to educate a daughter or a son is influenced by other factors wecontrol for in the regressions. See Section 3 for a more comprehensive description.

27Financial problems and opposition from the Church doomed the success of the 1842 education bill. Although

11

the municipal level, as they historically provided education and could have fostered human

capital in the area.28,29 Information includes monasteries’ foundation and dissolution dates, the

order to which they belonged and whether the community was masculine or feminine. We limit

our observations to monasteries founded before 1800. Although both beguines and traditional

nuns were devoted to the study of the scriptures, the latter were more inclined to a cloistered,

contemplative life, and therefore were less integrated with the secular society.

Secondly, we introduce the geographical distance from each municipality to Leuven, which

hosted the unique university in Belgium from its constitution in 1425 until 1817 when Liege

and Ghent universities were established. Being close to the unique university might have raised

awareness of the benefits of education or it may have intensified the local supply of books,

teachers and professors.30

Other Controls. In a pre-modern society, the degree of economic development is likely

to exert an influence on literacy levels and gender roles. To take into account the extent of

industrialisation, we introduce the number of steam engines per capita at the municipal level

using data from van Neck (1979, p.468-476). We incorporate additional variables obtained using

GIS solutions: latitude, longitude, distance to the closest river31 and to the closest large city.32

The proximity to waterways and urban areas is a predictor of industrialisation, and we argue

that the presence of close-by factories may affect the incentives on whether to educate children.

During the 19th century, agriculture was an important activity, and we introduce measures

to capture its extents. In particular, we include agricultural caloric yields, soil types and crop-

specific yield for staple crops: barley, buckwheat, rye, wheat, and white potatoes. The economic

literature indicates that historical agricultural productivity, measured by its caloric yield, has

shaped gender roles through economic specialisation. On the other hand, contemporaneous yield

levels affect the marginal productivity of labour, and hence the returns to education. For these

reasons, the variable takes on different values before and after 1550 (Galor and Ozak (2016)).

the reform contributed to increasing literacy rates, by 1880 39% of the population was still illiterate while thisfigure was as high as 51% in 1843.

28Information on monasteries has been derived from Cottineau (1939), Berliere (1955), Michel (1923) andDepartment of History, The Ohio State University (2016). Monasteries were assigned to the relevant municipalityexisting in 1866. For instance, the priory of Oignies is located in the town of Aiseau although its name suggests itwas in Oignes. Most religious institutions were suppressed, and their properties auctioned after the Napoleonicinvasion.

29Unfortunately, contemporaneous information on schools is only available at the provincial level.30The presence of this university should not directly encourage women to pursue higher education because

they were not admitted until 1880, first at Universite Libre de Bruxelles. Therefore, we expect distance to Leuvento have a negative impact on our measures of compared literacy, if any.

31We use shape-files from European Environment Agency (2017) tracing all major European rivers.32The 50 most populated cities are classified as large municipalities.

12

Carranza (2014) documents that soil types determine female employment in agriculture, and

a long tradition of female participation in agriculture can generate persistence. Following her

intuition, we compute the share of clayey, silty or sandy soil in a 10 km-radius area around each

municipality. Similarly, crop-specific yield measures from FAO’s Global Agro-Ecological Zones

(GAEZ) account for the possibility that women had a comparative advantage in given crops.33

Table 1 provides the summary statistics for the main variables.

4 Regression Analysis

Our baseline econometric specification is an OLS model, where the unit of observation is a

Belgian municipality existing in 1866:

outcomei(j) = α beguinagei(j) + βXi(j) + γj + εi(j). (6)

outcomei(j) measures gender equality and is observed for municipality i belonging to canton

j, beguinagei(j) is a variable attesting the historical presence of a beguinage, Xi(j) is a set of

covariates, γj are fixed-effects at the canton level, and εi(j) denotes the error term.34 As detailed

in Section 3, we proxy gender equality using mid-19th-century literacy differentials between

genders derived from census data.

We argue that beguinages offered marriageable girls an additional option besides marriage or

monastic life. In turn, access to a broader choice-set raised their bargaining power, allowing

them to improve their status in the society.35 As described in Section 5, we postulate that the

ensuing more gender-egalitarian view of the society is transmitted horizontally across generations

(as in Bisin and Verdier (2001)), fossilising in a local culture of higher equality between genders.

In that sense, children in towns with a beguinage were first exposed to beguines; and second, to

peers who were becoming more favourable towards women.

Table 2 presents the results of increasingly demanding OLS regressions. Columns 1) and 2)

present the raw correlation between the presence of beguinages and literacy equality indexi.36

33Crop yield data is provided by FAO/IIASA (2011). These measure production capacity for several cropsbased on geographic and climatic characteristics. In our setting, we consider low-input and rain-fed cultivation,mimicking medieval characteristics; and intermediate level of inputs and the use of irrigation to better account for19th-century agricultural techniques.

34Cantons are the administrative level immediately above municipalities. Belgium counted 193 cantons in 1886and each contained between one and 40 municipalities.

35Complementing this channel, beguines acted as role modellers by successfully and independently conductingtheir own lives without the intervention of men, which signalled their merit and equal role with respect tomales. On top of pious donations, beguinal independence was sustained by means of work —including high-level,prestigious professions as teachers—. Moreover, their occupations contributed to tightening the relationshipbetween beguines and townspeople.

36Appendix B presents the results when the dependent variable is female literacy sharei(j) and

13

Table 1: Summary statistics.

Mean Std.Dev. Min. Max.

Beguinage presenceBeguinage (0/1) 0.026 0.159 0 1Intensity: No Beguinage 0.974 0.159 0 1Intensity: 1 Beg., < 200 years 0.007 0.086 0 1Intensity: 1 Beg., > 200 years 0.012 0.108 0 1Intensity: > 1 Beg., > 200 years 0.003 0.054 0 1Intensity: > 3 Beg., > 200 years 0.004 0.061 0 1Exposure (centuries) 0.134 1.065 0.000 22.440

OutcomesLiteracy equality index, 1866 0.822 0.137 0.236 1.808Female literacy share, 1866 0.448 0.042 0.191 0.644Female literacy index, 1866 0.856 0.122 0.256 1.601

ControlsTotal men, 1866 (thousands) 0.949 2.622 0.009 74.169Total women, 1866 (thousands) 0.944 2.909 0.009 83.736Nuptiality men, 1866 0.360 0.036 0.181 0.669Nuptiality women, 1866 0.398 0.037 0.202 0.626Fem. monastery 0.030 0.184 0 2Masc. monastery 0.024 0.170 0 3Other monastery 0.072 0.259 0 1Distance to river (km) 9.082 8.757 0.002 52.396Distance to Leuven (km) 69.560 33.467 0.000 167.249Distance to beguinage (km) 16.265 18.164 0.000 122.010Distance to large city (km) 18.577 19.988 0.000 114.328Steam engines per 1000 people 0.071 0.689 0.000 12.493Potential caloric yield before 1550 2142.122 72.760 1908.810 2305.816Potential caloric yield after 1550 8894.732 310.662 8292.416 9780.832Potatoes, yield; pre-1550 3595.284 228.123 2726.000 3933.000Potatoes yield; post-1550 7206.351 450.825 5698.000 7724.000Wheat yield; pre-1550 2507.358 73.550 2328.000 2687.000Wheat yield; post-1550 5992.998 239.711 5163.000 6353.000Barley yield; pre-1550 2507.358 73.550 2328.000 2687.000Barley yield; post-1550 5992.998 239.711 5163.000 6353.000Rye yield; pre-1550 1545.315 57.366 1337.000 1735.000Rye yield; post-1550 3368.416 96.500 2972.000 3637.000Buckwheat yield; pre-1550 1032.003 38.934 911.000 1122.000Buckwheat yield; post-1550 1597.474 55.684 1486.000 1734.000

Observations 2553

14

Columns 3) and 4) introduce canton fixed-effects as an additional regressor, tackling regional

variation in attitudes towards women. In Columns 5) and 6), exogenous variables related to

geography are incorporated, namely, latitude, longitude, distance to Leuven, to the sea and

to the closest river, caloric yield, soil types, crop-specific productivity and the presence of

monasteries. Finally, Columns 7) and 8) incorporate demographic variables measured in 1866,

including the number of men and women, marriage rate, sex ratio, the percentage of domestic

and international migrants for each gender and an indicator variable for the 10% most populous

municipalities.37 In all specifications, we provide Conley standard errors in addition to standard

errors clustered at the canton level. The former control for spatial correlation and we use a 20

km bandwidth together with a Bartlett kernel.

Results show that municipalities where medieval beguine communities established displayed

higher levels of gender equality in 1866 when the institution already decayed or hosted only a

few women if any. The association remains valid across all specifications and regardless of how

the presence of beguinages is measured. In the most comprehensive specification, the presence of

a beguine community increases the variable literacy equality indexi(j) by 4.4 percentage points,

that is, 5% of its average value. The size of the effect we estimate is meaningful and equivalent

to a 2% increase in the number of literate women. Alternatively, increasing the presence of

beguinages by one standard deviation raises gender equality approximately as much as moving

one standard deviation away from the city of Leuven, which hosted a university promoting

education only for men, thus deflating our outcomes of interest. It should be noted that this

result is based on the regression reported in Column 7) of Table 2 where some controls are

potentially outcome variables. This is suggested by the drop in coefficients of interest in Columns

7) and 8) while before they were relatively stable. If that were the case, the effect of the

presence of beguinages would be greater, meaningfully raising literacy equality indexi(j) by 12.5

percentage points, equal to 15% of its average value, according to Column 5). Either way, a

longer presence of beguines is always related to increased female agency.

We argue that the presence of a beguinage in a town increased girls’ bargaining power by

improving their outside options. We test this mechanism exploiting the presence of feminine

monasteries because these also enlarged women’s choice-set. However, we stress that entering a

traditional convent conveyed less freedom and benefits than joining a beguinage. In particular,

female literacy indexi(j). These are similar to those presented in Table 2 and indicate a strong associationbetween beguinages and gender equality.

37Note that these variables may depend on the presence of beguinages.

15

Table 2: Beguinages and literacy equality.

Dep. variable: Lit. equality index, 1866

Baseline Fixed-effects Geography All

(1) (2) (3) (4) (5) (6) (7) (8)

Beguinage (0/1) 0.144 0.153 0.128 0.044(0.019)∗∗∗ (0.019)∗∗∗ (0.019)∗∗∗ (0.013)∗∗∗

[0.018]∗∗∗ [0.017]∗∗∗ [0.016]∗∗∗ [0.011]∗∗∗

Exposure (centuries) 0.021 0.022 0.017 0.006(0.003)∗∗∗ (0.003)∗∗∗ (0.003)∗∗∗ (0.003)∗∗

[0.003]∗∗∗ [0.003]∗∗∗ [0.003]∗∗∗ [0.003]∗∗

Fixed-effects No No Canton Canton Canton Canton Canton CantonGeography No No No No Yes Yes Yes YesDemography No No No No No No Yes YesObservations 2549 2549 2549 2549 2549 2549 2549 2549R2 0.030 0.029 0.203 0.200 0.219 0.214 0.433 0.432

Note: This table presents the results of OLS regressions relating the presence of beguinages to increased gender equality proxied by the variable literacy equality indexi(j ,see main text for a description. Columns 1) and 2) include only the presence of beguinages as a regressor. Columns 3) and 4) add fixed-effects at the canton level. InColumns 5) and 6) exogenous variables are incorporated, namely, the presence of monasteries, latitude, longitude, distance to Leuven, to the sea and the to closest river,caloric yield, soil types and crop productivity. Finally, Columns 7) and 8) incorporate demographic variables measured in 1866, including, the number of men and women,nuptiality, sex ratio, the share of domestic and international migrants and a dummy for the 10% most populous municipalities. Standard errors clustered at the canton levelin parenthesis and Conley standard errors, with a Bartlett kernel and a cut-off distance of 20 km in brackets. ∗ p < 0.1, ∗∗ p < 0.05, ∗∗∗ p < 0.01.

16

Table 3 assesses whether gender equality was larger in towns with a monastery for women,

for the sub-sample of towns without a beguinage. Column 1) considers feminine monasteries

altogether, while Column 2) distinguishes between open and cloistered ones according to the

Catholic order they belonged to.38 Regressions exclude municipalities that hosted a beguinage.

Based on the results, the presence of feminine monasteries increased gender equality, more

so for religious orders that interacted with the population. Arguably, open monasteries offered

girls an alternative option besides marriage while cloistered life indicates sheer vocation. Thus,

our results point to a more general pattern: the availability of a non-stigmatised environment

for young marriageable girls —either religious or semi-religious— that contributed to the

improvement of women’s status in the local society.39 Moreover, results preclude an alternative

interpretation based on monasteries only promoting literacy, insofar masculine monasteries did

so as well.40

4.1 Robustness

Beguinage location itself may be explained by town-specific attitudes regarding women, in

particular, a local culture accepting beguines’ lifestyle. In that sense, previous correlations

would only indicate the perpetuation of these characteristics, and the mechanism we ascribe

to beguinages would become suspect. We conduct a series of robustness checks to alleviate

this concern, before turning to an IV approach later. These more demanding specifications

reinforce the previous findings, lending additional credence to our hypothesis. In the interest

of readability, we report only the results of regressions including the largest set of regressors

(i.e. specifications control for geographical and agricultural characteristics, demographic and

development indicators as well as canton fixed-effects).41

First, Table 4 restricts the sample to municipalities with a similar culture and attitudes

towards women by selecting only towns located at close distance from a beguinage: 5 km, 10 km

and 20 km, in Columns 1) and 2), 3) and 4), and 5) and 6), respectively. The underlying idea is

that those municipalities near the ones chosen by beguines did not hold different views on gender

roles ex-ante. In other words, municipalities sufficiently close can be considered substitutes in

38 Cloistered orders include the Annoncides, Carthusian, Colettine, Dominicans and Trappists.39Nowadays, similar networks of women leaving their families to find mutual protection and support in other

women are present in a few developing societies. These groups are described in Appendix A.40Using the entire sample and further controlling for beguinage presence does not change the results.41As before, removing variables that are potential outcomes from regressors delivers larger estimates for the

presence of beguinages. Hence, our baseline coefficients can be seen as lower bounds of the estimation of interest.

17

Table 3: Literacy and gender equality; monasteries.

Dep. variable: Lit. eq. index, 1866

(1) (2)

Fem. monastery 0.047(0.019)∗∗

[0.018]∗∗∗

Fem. monasteryNo monastery Ref.1 open 0.047

(0.024)∗∗

[0.022]∗∗

2 open 0.306(0.008)∗∗∗

[0.016]∗∗∗

1 closed 0.039(0.059)[0.053]

1 unknown 0.011(0.029)[0.028]

Masc. monastery −0.010 −0.008(0.013) (0.013)[0.012] [0.012]

Other monastery 0.012 0.013(0.019) (0.019)[0.017] [0.018]

Fixed-effects Canton CantonGeography Yes YesDemography Yes Yes

Observations 2479 2479R2 0.407 0.407

Note: This table presents the results of OLS regressions relating the presence of monasteries to increasedgender equality. It documents higher levels of gender equality in towns that offered more alternatives to girls.See main text for a discussion of the mechanism. Column 1) introduces the number of monasteries for eachgender while Column 2) distinguishes, among feminine monasteries, those that belonged to open or enclosedCatholic orders. “No monasteries present” is the omitted reference category. Geography: latitude, longitude,distance to Leuven, to the sea and rivers, caloric yield, soil types and crop productivity. Demography: numberof men and women, nuptiality, sex ratio, the share of internal and external migrants and a dummy for the10% most populous municipalities. Regressions include canton fixed-effects. Municipalities that hosted abeguine community are excluded from the sample. Standard errors clustered at the canton level in parenthesisand Conley standard errors, with a Bartlett kernel and a cut-off distance of 20 km in brackets. ∗ p < 0.1,∗∗ p < 0.05, ∗∗∗ p < 0.01.

18

Table 4: Beguinages and literacy equality; sub-samples by distance.

Dep. variable: Lit. equality index, 1866

5Km 10Km 20Km

(1) (2) (3) (4) (5) (6)

Beguinage (0/1) 0.040 0.042 0.047(0.020)∗∗ (0.015)∗∗∗ (0.013)∗∗∗

[0.015]∗∗∗ [0.013]∗∗∗ [0.012]∗∗∗

Exposure (centuries) 0.007 0.007 0.007(0.004)∗ (0.003)∗∗ (0.003)∗∗

[0.003]∗∗ [0.003]∗∗∗ [0.003]∗∗

Fixed-effects Canton Canton Canton Canton Canton CantonGeography Yes Yes Yes Yes Yes YesDemography Yes Yes Yes Yes Yes YesObservations 371 371 1114 1114 2060 2060R2 0.664 0.664 0.518 0.517 0.436 0.435

Note: This table presents the results of OLS regressions relating the presence of beguinages to increased genderequality. Columns 1) and 2) consider only municipalities located less than 5 km from a beguinage, 3) and 4)increase the distance up to 10 km and 5) and 6) further expand it to 20 km. Geography: latitude, longitude,distance to Leuven, the sea and rivers, caloric yield, soil type and crop productivity. Demography: population,nuptiality and %age of domestic and international migrants by gender, sex ratio and a dummy for the 10%most populous towns. Regressions include canton fixed-effects. Standard errors clustered at the canton level inparenthesis and Conley standard errors, with a Bartlett kernel and a cut-off distance of 20 km in brackets.∗ p < 0.1, ∗∗ p < 0.05, ∗∗∗ p < 0.01.

terms of attitudes towards women. Overall, results indicate a positive and significant effect of

beguine communities on gender equality. Moreover, the estimated coefficients are stable. Further

adding beguinage fixed-effects and clustering the standard errors at that level does not change

the results.42

Second, in Table 5, we use distance to the closest beguinage as the main regressor. The idea

is to depart from indicators exhibiting numerous zeros such as beguinagei(j) (about 3% of the

municipalities ever had a beguinage). Distance to the closest beguinage, measured in log-km,

solves this issue as it varies continuously. Related, we follow Shiode and Shiode (2011) and

exploit towns’ network structure and define an Inverse Distance Weighting indicator of exposure

to the historical beguinal presence (IDW), relying on the distance decay effect. It is based on the

simple notion that the interaction between two communities declines as the distance between

them increases. The IDWi indicator summarises the influence of any known beguinage on a

given municipality, with an effect that is inversely proportional to its distance. This measure

effectively considers the town’s relative isolation in the network. More in detail, it is calculated

42These results are available from the authors upon request.

19

as follows: IDWi(j) = log(∑

k1/Dist(i(j),k(j))1(Beguinagek(j))∑

k1/Dist(i(j),k(j))

). The adoption of this indicator as

the independent variable contributes to our identification in two ways. First, it corrects for

the endogeneity in the location choice. The identification assumption, in this case, is that, for

municipality i(j), the presence of a beguinage in another municipality j is uncorrelated with

unobserved determinants of gender norms. Second, it captures spillover effects: even in towns

not hosting a beguine community, girls had the opportunity of moving elsewhere to enter the

movement. Besides using the indicator variable 1(Beguinagek(j)), we also weigh observations

by exposurei(j) to account for the availability of beguinages over time.

Table 5: Literacy and gender equality; distance.

Dep. variable: Lit. equality index, 1866

(1) (2) (3)

Min. Distance Beg. (log-km) −0.015(0.005)∗∗∗

[0.005]∗∗∗

IDW 0.085(0.024)∗∗∗

[0.022]∗∗∗

IDW, exposure 0.103(0.029)∗∗∗

[0.025]∗∗∗

Fixed-effects Canton Canton CantonGeography Yes Yes YesDemography Yes Yes YesObservations 2549 2475 2475R2 0.432 0.438 0.438

Note: This table presents the results of OLS regressions relating the presence of beguinages to increased genderequality proxied by the variable literacy equality indexi(j), see main text for a description. Column 1) employsdistance to the closest beguinage in log-km, Column 2) introduces IDW as regressor while Column 3) usesthe IDW weighted using exposure. Geography: latitude, longitude, distance to Leuven, the sea and rivers,caloric yield, soil type and crop productivity. Demography: population, nuptiality and %age of domestic andinternational migrants by gender, sex ratio and a dummy for the 10% most populous towns. Regressionsinclude canton fixed-effects. Standard errors clustered at the canton level in parenthesis and Conley standarderrors, with a Bartlett kernel and a cut-off distance of 20 km in brackets. ∗ p < 0.1, ∗∗ p < 0.05, ∗∗∗ p < 0.01.

The results of this exercise are in line with our previous findings: being further away from a

beguinage reduces gender equality. Moreover, the results we obtain when using IDWi(j) indicate

higher gender equality in towns better connected to beguine municipalities. In that sense, a

closer location to clusters of beguinages offered girls multiple alternatives to avoid marriage,

presenting them even a broader choice set compared to a single beguinage. This raises gender

equality levels. Also, Table 4 only captures the marrying opportunity cost effect for girls and

20

not any role modelling effect.

In Appendix B, we conduct a series of additional robustness checks that further validate

the uncovered relation by altering the definition of the treatment and control sample and

considering alternative independent variables. First, Table 8 introduces alternative measures

of beguinage presence: a new variable accounting for the presence of beguinages combining

features of beguinagei and exposurei, and a modified version of exposurei(j) not considering

time-overlapping beguinages. Second, we control for the sensitivity of the variable IDWi(j), by

proposing an alternative measure, although the resulting estimates are in line with the baseline

ones. In particular, removing towns with a beguinage from the sample provides similar results.

Furthermore, assuming disconnected networks between linguistic communities yields equivalent

results. Third, at the time of the census, there were few beguinages operative. We remove

such towns from the sample to capture the cultural legacy of these communities and not the

contemporaneous effect they exerted on the local population. We also introduce male literacy

rates as regressor and further control for municipal charters, the instrument we use later. We

follow Valencia Caicedo (2018) and compare the effect of beguinages that were abandoned soon

after they were established to those that remained. It can be argued that any town beguines

selected had similar attitudes towards gender equality. If our estimates only captured this aspect

and not social changes induced by the presence of beguines, shortly-lived beguinages should

influence gender attitudes. We do not find evidence of that as Table 12 indicates. Furthermore,

we estimate the effect of beguinages on equality using propensity score matching. Supplementing

these robustness checks, we generate placebo beguinages to dismiss concerns arising from the

low number of beguinages present in the sample. For this purpose, the sample is divided, using

total population, in quantiles and the expected number of beguinage is calculated for each.

A municipality in each quantile receives a placebo beguinage with probability equal to the

quantile-specific expected number of beguinages. Based on 100 repetitions, we find that in

less than 15% of the cases placebo beguinages are significant. Finally, following Oster (2016)

we estimate the strength of selection on unobservables necessary to diminish the estimated

parameters of interest down to zero, when using the limited and full set of controls. In our case,

selection on unobservables should be larger than on observables for the limited set of controls,

considering an increase in the potential R2 of 30%. When the full set of controls is included,

results indicate possible selection on unobservables only when the main regressor is exposure

and when the potential R2 is increased above 25% for Beg.(0/1).

21

Instrumental Variable Approach. In order to better address the potential endogeneity

of beguinage location, we instrument the historical presence of beguines exploiting exogenous

changes in the local organisation of power that occurred before the advent of the movement.

Considering that beguines supported themselves working in several crafts, teaching and trading,

locations facilitating market access should have attracted these communities more. The privilege

to set-up markets and guilds, among others, were typically granted in municipal charters,

documents conveying additional rights to selected municipalities. Indeed, in the medieval

European context, some towns were awarded a semi-independent status through the acquisition

of new privileges encoded in charters. A municipality receiving a charter was granted partial

exemption from wars for its citizens, often allowing the opportunity to wall the town, trade,

mint currency, set up a municipal judicial system, and establish guilds. Municipal charters, in

fact, decentralised decision-making.43

This change in the local powers and structures could be achieved for various motives. Some

lords founded villages and immediately endowed them with a municipal charter.44 Other towns

benefited from their strategic position and obtained a municipal charter in exchange for defence.45

A few Lords sought increasing trade and urban development and granted charters to this end.46

Finally, some towns gained this privileged status by siding with a Lord at times of conflict

or because of specific local circumstances.47 In any case, granting a municipal charter was a

prerogative of the ruler: not all cities that ever requested this special status obtained it.

In light of the benefits conveyed by municipal charters in terms of trade and economic

activity in general, and considering beguines’ occupations, beguinages were more likely to

establish in towns already enjoying a municipal charter. Moreover, the more decentralised

43The majority of these documents simply transposed privileges already bestowed to other municipalities,effectively copying them. In most cases, Belgian municipalities celebrated their independence erecting a belfry :a monumental, fortified building that served as a court, prison, archive and treasure chamber and doubled aswatching and bell tower. Eastern cities under the control of the Prince-Bishop of Liege constructed, instead,small, decorated columns named perron. Interestingly, when the city of Liege capitulated in 1467 its perron wasdismantled and moved to Bruges for ten years. The peace treaty remarked that “the Duke wanted nothing leftstanding from the ancient Liege constitution, nor from public liberties, the legacy of a distant past”.

44This is the case of Herentals, Hoogstraten and Landen.45Ath, Geraardsbergen, Maaseik, Vilvoorde, Stokkem and Zoutleeuw received charters for being border posts

against powerful regional rivals.46Joan and Margaret II, Countesses of Flanders, Henry I, Duke of Brabant, and Philip I, Count of Flanders,

were among the group of nobles who promoted these reforms. Their reformation spirit benefited the cities ofAntwerp, Damme, Eeklo, Landen, Oostende and Roeselare.

47Bruges and Kortrijk were involved in a local conflict confronting Thierry of Alsace and William Clito. TheLord of Ghent delegated the resolution of internal turmoil to the citizens by issuing a charter. Phillip II of Francebought off Tournai’s inhabitants offering them a municipal charter. As an example of a more peculiar motivation,the Prince-Bishop of Liege granted Huy the first Belgian charter to reward the city for its involvement in thereconstruction of the cathedral. Abbot Leonius interceded with Thierry, count of Flanders and a friend of his, togrant a municipal charter to his vassals of Poperinge.

22

administrative framework of chartered towns made it easier for beguines to obtain the terrain to

erect their communal buildings, sometimes also as a donation. In view of these characteristics,

municipal charters are employed as a dummy instrument to predict locations that beguines

chose to establish communities. In order to have a clear identification strategy, we consider only

municipal charters that were bestowed before the 13th century, as beguinages were only founded

after this time. Imposing this restriction assures that beguinages did not exert any influence in

obtaining a municipal charter. In addition, the choice of pre-13th-century municipal charters

rules out towns seeking a charter with the goal of attracting beguines. In total, 42 Belgian

municipalities were granted a municipal charter before the 13th century.48 Among the 70 towns

with a beguinage, 32 held a charter before beguines settled.

Since trade and overall economic activity historically led to population growth and increased

returns to human capital, municipal charters may have promoted female literacy beyond operating

exclusively through the presence of beguine communities. In order to strengthen our exclusion

restriction, we emphasise that the outcome of interest is not literacy in absolute terms but

gender equality in this dimension. In this regard, we see female-to-male literacy measures as

indicators of female agency. Moreover, if municipal charters boosted the demand for schooling,

this alleged higher education would bias our estimates downward, as men were traditionally

favoured over women in education investments. We can also dismiss that municipalities that

hosted a beguinage grew faster.49

Additionally, we argue that towns that never hosted a beguinage were comparable in terms

of acceptance of female independence. Similarly, places where beguines settled should be

comparable in terms of gender equality. In the Appendix, supplementary estimates measure the

effect of having received a municipal charter separately for towns with and without a beguine

community.50 If, indeed, municipal charters directly promoted female literacy, municipalities

that received a charter should score higher in female-to-male literacy on this sample. Results in

Column 1) do not support this theory. We repeat the same procedure focusing on municipalities

that never hosted a beguinage in Column 2), obtaining similar conclusions. Finally, Panel C)

of Table 11 in the Appendix includes municipal charters as regressor, showing that municipal

48These are: Aalst, Aarschot, Antwerp, Ath, Beringen, Borgloon, Brugge, Bruxelles, Damme, Deinze, Den-dermonde, Diest, Diksmuide, Eeklo, Ghent, Geraardsbergen, Gistel, Halen, Halle, Harelbeke, Hasselt, Herentals,Hoogstraten, Huy, Kortrijk, Landen, Leuven, Lier, Maaseik, Ninove, Oostende, Oudenaarde, Poperinge, Roeselare,Sint-Truiden, Soignies, Stokkem, Thurnhout, Tournai, Vilvoorde, Ypres and Zoutleeuw.

49We use data from Cuvelier (1912) on municipal size in 1438 and compute population growth rate betweenthis date and 1886. Unfortunately, the geographical coverage is limited to the historical province of the Brabant.

50See Table 13.

23

charter did not contribute to gender equality.

IV Results. Table 6 presents the findings of the IV approach when beguinage location

is instrumented using municipal charters. In Columns 1) and 2), the dependent variable

is literacy equality indexi(j), Columns 3) and 4) focus on female literacy sharei(j) and the

remaining use female literacy indexi(j). In general, F-statistics are well above 10, indicating

that regressions are not biased due to the inclusion of weak instruments. IV estimations follow

Wooldridge (2010) procedure 21.1 that correctly accounts for the non-linearity of the first-stage

when the independent variable of interest is beguinagei.51 This method differs from standard

2SLS by introducing the estimated value of the endogenous variable as an instrument. The IV

procedure for exposurei(j) follows the conventional two-step estimator.

Table 6: Beguinages and Literacy equality, IV.

Lit. eq. index, 1866 Fem. lit. share, 1866 Fem. lit. index, 1866

(1) (2) (3) (4) (5) (6)

Beguinage (0/1) 0.075∗∗∗ 0.018∗∗∗ 0.069∗∗∗

(0.019) (0.006) (0.019)Exposure (cent.) 0.013∗∗ 0.004∗∗ 0.013∗∗

(0.005) (0.002) (0.005)

Fixed-effects Canton Canton Canton Canton Canton CantonGeography Yes Yes Yes Yes Yes YesDemography Yes Yes Yes Yes Yes Yes1st-st. F-val. 54.9 28.4 54.9 28.4 54.9 28.4Mun. charter 2.644∗∗∗ 3.103∗∗∗ 2.644∗∗∗ 3.103∗∗∗ 2.644∗∗∗ 3.103∗∗∗

(0.586) (0.121) (0.586) (0.121) (0.586) (0.121)

Observations 973 2549 973 2549 973 2549R2 0.470 0.428 0.448 0.406 0.264 0.228

Note: This table presents the results of IV regressions relating the presence of beguinages to increased genderequality where the presence of beguinal communities is instrumented using municipal charters. Geography:latitude, longitude, distance to Leuven, to the sea and to rivers, caloric yield, soil types and crop productivity.Demography: the number of men and women, nuptiality, sex ratio, the percentage of domestic and internationalmigrants and a dummy for the 10% most populous municipalities. Standard errors clustered at the cantonlevel. ∗ p < 0.1, ∗∗ p < 0.05, ∗∗∗ p < 0.01.

Overall, results from the instrumented specification indicate that beguinages had a positive,

significant and enduring effect on gender equality. Moreover, even if our approach cannot assuage

all the possible endogeneity concerns, this exercise is useful insofar it highlights that beguine

communities were instrumental in maintaining a higher regard for women for over 600 years.

However, a clearer identification strategy limiting the IV approach only to cases where obtaining

51In this case, the first-stage follows a probit model. Columns 1), 3) and 5) of Table 6 report this statistic.

24

a municipal charter was arguably more exogenous delivers similar results.52

In Appendix B, we provide multiple robustness tests of the IV specification that further

confirm previous findings and generally strengthen the results. These are the same, when

applicable, to those performed for the OLS analysis.

5 A Model of Cultural Transmission

So far, our research unveiled an enduring relationship between beguine communities and gender

equality. Possible explanations include beguines lifting girls’ literacy rates as teachers, role-

modelling effects and increased female bargaining power. However, the previous Sections evince

that neither role-modelling nor beguinal teaching efforts caused the increase in gender equality

we measure.

Consequently, we propose a theoretical model to explain our results emphasising girls’

bargaining power as the relevant mechanism at play. It illustrates how beguinages triggered

a shift in values that could spread beyond the beguinal population itself, influencing gender

outcomes for centuries. We believe this captures the essence of what beguinages offered to young

girls, while other potential mechanisms are left unexplored.

In this framework, cultural change in gender roles operates through the marriage market. We

consider a population of men —denoted by m— and women —f— in equal number who randomly

meet in the local marriage market. Before getting married, they bargain non-cooperatively over

the share of the constant marital surplus generated by a union, denoted by y.53 Let sm,ft ∈ (0, 1)

be the couple-specific share men appropriate and t = 1, 2, .. denote the time period. Agents will

agree to the wedding if they enjoy their share of marital surplus more than their life as single.

Individuals are characterised by their outside option in case they do not marry. Women f are

assumed to be homogeneous in their outside option uf ∈ (0, y) and we posit that heterogeneous

men m follow a Type I Pareto distribution with scale parameter xt and shape parameter α = 2

such that um,t ∼ ParetoI(xt, 2).54 The initial average type of men is: µ0 = αx0α−1 .

52Reasons considered include charters bestowed upon newly founded cities, conflict, and other miscellaneousinstances. The result of this more demanding specification validates our previous findings. However, the estimatedcoefficient becomes less precise because the instrument is able to predict fewer beguine communities. These resultsare not reported but are available upon request.

53Many aspects are typically encompassed in the notion of marital surplus: children and other household publicgoods, but also other psychological and sociological amenities, as self-realisation and social recognition.

54A Type I Pareto distribution with value 2 for the shape parameter allows obtaining closed-form solutions.

25

The potential couples’ negotiation problem follows Nash bargaining:

maxsm,ft

(sm,ft y − um, t

)β((1− sm,ft

)y − uf

)1−β, (7)

where β ∈ (0, 1) denotes men’s bargaining power and is assumed to be constant.

At the optimum, men who marry enjoy the share sm,ft

?=

um,t(1−β)+β(y−uf )y , which is

decreasing in uf Matched individuals for which um,t + uf > y cannot agree on a sharing rule.

Consequently, men at the right tail of the distribution will stay single. Both very demanding

men and the women matched with them are better off if they do not marry.

Each married couple has one daughter and one son. To obtain a tractable model, daughters

inherit the constant type uf of their mothers while sons are horizontally socialised. In particular,

sons’ type follows a Type I Pareto distribution with α = 2 and the average son type equals the

average share of surplus that married men obtain in the bargaining process. In other words,

following Bisin and Verdier (2001), a son observes married couples to infer the minimum amount

he should demand in the bargaining process. This average equals:

µt+1 = E(sm,ft

?y)

= y

∫ y−uf

0sm,ft

?f(um,t)dum,t∫ y−uf

0f(um,t)dum,t

= β(y − uf ) + (1− β)2xt(y − uf )

xt − uf + y

Since the average of a Type I Pareto distribution is given by αx1−α we can rewrite the previous

expression in terms of µt55:

µt+1 = (y − uf )

(β +

2(β − 1)µt−µt + 2uf − 2y

). (8)

Equation 8 describes the dynamic evolution of µ over generations. For sufficiently large

values of the initial µ0, µt+1 < µt∀t.56

In order to study the long-term evolution of gender norms, we focus on the change of men’s

values over generations, as captured by µ. We analyse the unique, asymptotically steady-state

55We assume α = 2 and µ0 < 2 ∗ (y − uf ). The latter is a technical condition to have a non-zero denominatorfor the expected average of the truncated distribution.

56More in detail, the condition on µ0 reads µ0 >12(uf − y)(β−

√β(8 + β)). This is not particularly restrictive

considering medieval gender norms according to which husbands were in charge of household administration andwere considered the legal guardians of their wives.

26

level of µt given by:57

limt→∞

µt = µ? =1

2(y − uf )

√β(√

8 + β −√β). (9)

Starting from this model, we discuss how the presence of a beguinage in the municipality may

have influenced the beliefs about the status of women. We maintain that becoming a beguine

was a competing alternative to marriage that broadened girls’ choice-set. In particular, women

in municipalities with a beguinage enjoyed an increased outside option (ubeguinagef > uf ) and

could bargain a better status within the couple if married. In this scenario, a larger share of

selfish men will remain single, leading to a selection of values in the married population and to a

lower steady-state level of µ in the long run. It can be shown that µ?beguinage < µ?.58

Figure 4 illustrates the evolution of the average attitude of men over time under two different

scenarios: a no-beguinage case (dashed line) and a case in which a beguinage is founded in

a municipality and dissolves after some time (solid line). After the closure of the beguinage,

women face the baseline outside option uf .

Figure 4: Evolution over Time.

2 3 4 5t

Average type of men

Beguinage

No beguinage

Note: This Figure displays the evolution of the average attitude of men under two different scenarios: ano-beguinage case (dashed line) and a case in which a beguinage is founded (solid line). assumes β = 0.65 andx = 5 for both cases and ubeguinage

f = 1.2 > uf = 1. The beguinage is operative for three periods.

The presence of a beguinage, even for a limited period of time, results in less male-egotistic

values in the society. While the beguinage operates, the values converge to a different, lower

steady-state and once the beguinage disappears, the values start diverging from that path,

57The speed of convergence to the steady-state accelerates with β. In that sense, higher bargaining power formen increases the speed with which the share of egotistic men is eroded over time.

58A possible extension of the model could consider that men’s bargaining power decreases with female’s outsideoption. This would produce a set of different steady-states whose exact value would depend on beguinage’sduration, assuming it closes at some point of time.

27

returning to the no-beguinage equilibrium. Such evolution is in line with our empirical observation

of more gender-equal values in municipalities that enjoyed a beguinal institution for some time.

6 Conclusion

This paper proposes a novel mechanism to explain the differences in gender equality, grounded

in the strengthening of women’s bargaining power and enabled by the broadening of their

alternatives to marriage. We argue that this improved position in the marital decisions positively

affected women’s outcomes, closing the gap with men’s. The intergenerational transmission of a

gender-egalitarian view of the society generated persistence. Our theory on the deep roots of

gender roles complements the existing economic literature that mainly emphasises physiological

differences.

We illustrate our mechanism focusing on the Belgian setting, explaining how medieval

beguine communities broadened the set of girls’ options besides marriage. Beguines were single

women independent of men, living a semi-religious life outside the recognised orders in self-

supporting communities. These unique characteristics allowed them to delay marriage and

offered an alternative to monastic seclusion amid a society that traditionally disapproved such

lifestyle. Specifically, we show that medieval beguinages —founded between 1207 and 1500—

were instrumental in generating higher local levels of gender equality in 1866. In particular, we

estimate a 5.3% reduction in the gender literacy gap, our baseline measure of gender equality.

We develop a theoretical framework to rationalise the uncovered relationship.

Our study elucidates the importance of female opportunity cost of getting married to

lessen gender inequalities and promote the diffusion of egalitarian roles. Female associations

—beguinages— are crucial in our setup: they provide the actionable threat if nuptial negotiation

turns unsuccessful, thus raising female’s bargaining power. Appendix A describes modern-day

associations that could trigger analogous societal changes. Other social arrangements that can

equally promote gender equality are the traditions of uxorolocality and neolocality. The pivotal,

common element is the provision of a safety-net for women that de facto allows them a higher

claim on the marital surplus by reducing their fear for recrimination. In that sense, the social

acceptance of a previously rebuked behaviour operates similarly, as it happened for women

joining the labour force after World War II.

28

References

Akcomak, I. S., Webbink, D., and Weel, B. (2016). Why Did the Netherlands Develop So Early?

The Legacy of the Brethren of the Common Life. The Economic Journal, 593(126):821–860.

Alesina, A. and Giuliano, P. (2010). The power of the family. Journal of Economic Growth,

15(2):93–125.

Alesina, A., Giuliano, P., and Nunn, N. (2013). On the Origins of Gender Roles: Women and

the Plough. The Quarterly Journal of Economics, 128(2):469–530.

Andersen, T. B., Bentzen, J., Dalgaard, C.-J., and Sharp, P. (2017). Pre-reformation Roots of

the Protestant Ethic. The Economic Journal, 127(604):1756–1793.

Becker, S. O., Boeckh, K., Hainz, C., and Woessmann, L. (2016). The Empire Is Dead, Long

Live the Empire! Long-Run Persistence of Trust and Corruption in the Bureaucracy. The

Economic Journal, 126(590):40–74.

Becker, S. O. and Woessmann, L. (2008). Luther and the girls: Religious denomination and

the female education gap in nineteenth-century Prussia. Scandinavian Journal of Economics,

110(4):777–805.

Berliere, U., editor (1955). Monasticon belge. Centre national de recherches d’histoire religieuse,

Liege.

Bertocchi, G. and Bozzano, M. (2019). Origins and implications of family structure across Italian

provinces in historical perspective. In Cliometrics of the Family, pages 121–147. Springer.

Bethune, J. (1883). Cartulaire du beguinage de Sainte-Elisabeth a Gand. De Zuttere.

Bisin, A. and Verdier, T. (2001). The Economics of Cultural Transmission and the Dynamics of

Preferences. Journal of Economic Theory, 97(2):298–319.

Bowie, F. (1990). Beguine Spirituality: Mystical Writings of Mechthild of Magdeburg, Beatrice

of Nazareth, and Hadewijch of Brabant. Crossroad.

Bratti, M., Deiana, C., Havari, E., Mazzarella, G., and Meroni, E. (2017). What Are You Voting

for? Refugee Reception Centres and Voting at the 2016 Italian Constitutional Referendum.

Technical Report 11060, IZA Institute of Labour Economics.

Cage, J. and Rueda, V. (2016). The Long-Term Effects of the Printing Press in Sub-Saharan

Africa. American Economic Journal: Applied Economics, 8(3):69–99.

Carmichael, S. G., de Pleijt, A., van Zanden, J. L., and De Moor, T. (2016). The European

marriage pattern and its measurement. The Journal of Economic History, 76(01):196–204.

29

Carranza, E. (2014). Soil Endowments, Female Labor Force Participation, and the Demographic

Deficit of Women in India. American Economic Journal: Applied Economics, 6(4):197–225.

Cottineau, L. H. (1939). Repertoire topo-bibliographique des abbayes et prieures. Protat freres,

Macon.

Cuvelier, J. (1912). Les denombrements de foyers en Brabant (XIVe-XVIe siecle). Table

onomastique. Kiessling, Bruxelles.

De Moor, T. (2014). Single, Safe, and Sorry? Explaining the Early Modern Beguine Movement

in the Low Countries. Journal of Family History, 39(1):3–21.

De Moor, T. and Van Zanden, J. L. (2010). Girl Power: the European Marriage Pattern and

Labour Markets in the North Sea Region in the Late Medieval and Early Modern Period. The

Economic History Review, 63(1):1–33.

Department of History, The Ohio State University (2016). Monastic Matrix. Accessed: 2017-02-

28.

Doepke, M. and Tertilt, M. (2009). Women’s Liberation: What’s in It for Men? Quarterly

Journal of Economics, 124(4):1541–1591.

European Environment Agency (2017). WISE Main Rivers, draft April 2012 ECRINS - EEA

hydrographic data set. http://projects.eionet.europa.eu/ecrins/library/relation_

ecrins/national_reference/wise-main-rivers-v1-draft-april-2012. Accessed: 2017-

02-27.

FAO/IIASA (2011). Global Agro-ecological Zones (GAEZ V3.0). gaez.fao.org. Accessed:

2017-10-24.

Fernandez, R. and Fogli, A. (2009). Culture: An Empirical Investigation of Beliefs, Work, and

Fertility. American Economic Journal: Macroeconomics, 1(1):146–177.

Fong, Y.-F., Guo, R., Ng, T., and Wong, K.-Y. (2018). Sworn Spinsters: An Economic Analysis

of Celibacy. Technical report, mimeo.

Fredericq, P. (1889). Corpus documentorum Inquisitionis haereticae pravitatis Neerlandicae;

verzameling van stukken betreffende de pauselijke en bisschoppelijke inquisitie in de Nederlanden.

Uitg. door Paul Fredericq [et al.]. Gent J. Vuylsteke.

Galor, O. and Ozak, Omer. (2016). The Agricultural Origins of Time Preference. American

Economic Review, 106(10):3064–3103.

Galor, O. and Weil, D. N. (1996). The Gender Gap, Fertility, and Growth. The American

Economic Review, 86(3):374–387.

30

Geybels, H. (2004). Vulgariter beghinae : eight centuries of Beguine history in the Low Countries.

Brepols, Turnhout.

Giuliano, P. (2018). Gender: An historical perspective. In Averett, S. L., Argys, L. M., and

Hoffman, S. D., editors, Oxford Handbook on the Economics of Women. Oxford University

Press.

Giuliano, P. and Nunn, N. (2017). Understanding Cultural Persistence and Change. Technical

report, National Bureau of Economic Research.

Gottlieb, J. and Robinson, A. L. (2016). The effects of matrilineality on gender differences in

political behavior across Africa. University of California, Berkeley, mimeo.

Grosjean, P. and Khattar, R. (2017). It’s raining men! Hallelujah? The long-run consequences

of male-biased sex ratios. The Review of Economic Studies.

Hansen, C. W., Jensen, P. S., and Skovsgaard, C. V. (2015). Modern gender roles and agricultural

history: the Neolithic inheritance. Journal of Economic Growth, 20(4):365–404.

Iyigun, M. and Walsh, R. P. (2007). Endogenous Gender Power, Household Labor Supply and

the Demographic Transition. Journal of Development Economics, 82(1):138–155.

Jayachandran, S. (2015). The Roots of Gender Inequality in Developing Countries. Annual

Review of Economics, 7(1):63–88.

Klasen, S. (2002). Low Schooling for Girls, Slower Growth for All? Cross-Country Evidence on

the Effect of Gender Inequality in Education on Economic Development. The World Bank

Economic Review, 16(3):345–373.

Klasen, S. and Lamanna, F. (2009). The Impact of Gender Inequality in Education and Em-

ployment on Economic Growth: New Evidence for a Panel of Countries. Feminist Economics,

15(3):91–132.

Lowes, S., Nunn, N., Robinson, J. A., and Weigel, J. L. (2017). The evolution of culture and

institutions: Evidence from the Kuba Kingdom. Econometrica, 85(4):1065–1091.

Lynch, K. A. (2003). Individuals, Families, and Communities in Europe, 1200-1800: the urban

foundations of Western society, volume 37. Cambridge University Press.

McDonnell, E. W. (1954). The Beguines and Beghards in Medieval Culture: With Special

Emphasis on the Belgian Scene. Rutgers University Press.

Michel, E. (1923). Abbayes et monasteres de Belgique. G. Van Oest. Cie.

Mommaers, P. and Dutton, E. M. (2004). Hadewijch: Writer, Beguine, Love Mystic. Peeters

Publishers.

31

Oster, E. (2016). Unobservable selection and coefficient stability: Theory and evidence. Journal

of Business & Economic Statistics, pages 1–18.

Ruyskensvelde, S. V. (2016). Wartime Schooling and Education Policy in the Second World

War: Catholic Education, Memory and the Government in Occupied Belgium. Springer.

Shiode, N. and Shiode, S. (2011). Street-level spatial interpolation using network-based IDW

and ordinary kriging. Transactions in GIS, 15(4):457–477.

Simons, W. (2010). Cities of Ladies: Beguine Communities in the Medieval Low Countries,

1200-1565. University of Pennsylvania Press.

Tur-Prats, A. (2018). Family types and intimate-partner violence: A historical perspective. The

Review of Economics and Statistics.

Valencia Caicedo, F. (2018). The Mission: Human Capital Transmission, Economic Persistence,

and Culture in South America. The Quarterly Journal of Economics.

van Neck, A. (1979). Les debuts de la machine a vapeur dans l’industrie belge, 1800-1850. Palais

des Academies.

Vander Schelden, J. e. H., editor (1862). Requete presentee par les Grandes-dames et Superieures

des deux Beguinages a messieurs les President et membres du Conseil communal de Gand. J.

et H. Vander Schelden,, Gand.

Voigtlander, N. and Voth, H.-J. (2012). Persecution perpetuated: The medieval origins of anti-

semitic violence in Nazi Germany. The Quarterly Journal of Economics, 127(3):1339–1392.

Waldinger, M. (2017). The long-run effects of missionary orders in Mexico. Journal of Develop-

ment Economics, 127:355–378.

Warnkonig, L. A. (1836). Histoire de la Flandre et de ses institutions civiles et politiques jusqu’a

l’annee 1305. Hayer.

Wooldridge, J. M. (2010). Econometric Analysis of Cross Section and Panel Data. The MIT

Press.

32

Appendices

A Analogous Institutions

Despite the fact that beguinages are indeed a phenomenon unique to the cities of the Low

Countries, in the recent past we can observe other large, female-only prime examples of lay

efforts to construct artificial families and communities within an urban setting (Lynch (2003)).

Nowadays, there exist other institutions resembling beguinal social customs: they are collective

arrangements formed by groups of women in order to overcome certain common problems by

setting rules regarding membership, and the use of the resources and services the group owns

collectively. Another key feature is the adoption of de facto celibacy beyond apparent religious

concerns.

One of these institutions is called Zishunu: an ancient custom that originated from indigenous

groups in southwestern China and consists of a group of women called Sworn Spinsters. Zishunu

literally means self-combed women and depicts the local customs of girls forming sisterhoods

who ritually pinned up their hair, and openly took elaborate vows not to marry and to remain

childless. Evidence of such practices can be found in local publications and prospered well into

the end of the 19th century, when almost all of them worked in silk reeling factories and, from

the 1930s, as domestic servants in Hong Kong and Singapore. Their vow of celibacy, beyond

apparent religious concerns, represents a commitment device of high-intensity labour supply

and therefore made their employers invest more in their human capital match them to more

important working positions (Fong et al. (2018)).

Another instance is the one of women seeking the solidarity and protection of other women to

escape misogynistic marriages. This is the case of a community in Tanzania that has embraced

an age-old tradition of heterosexual women marrying women. In fact, in the Kurya community in

Tarime District of the Mara region in northern Tanzania, the so-called Nyumba Nthobu tribal law

is used by women who are not able to inherit property due to patriarchal cultural constraints.59

Marriage between the two women is justified by the bride price paid before marriage, since there

is no intimate relationship between the two women. In some cases, Nyumba Ntobhu can be a

polygamous marriage, as the older woman marries two younger women. This custom enables

them to claim the children born by the other woman60 as their own and is a way of providing

59Barren women or old women who do not have sons to inherit their property and whose daughters have movedaway to their husbands’ villages cannot take over property in the line of succession.

60A man is typically selected from the clan of the old woman for reproduction purposes with the young woman.

33

security for their old age. The culture is also being used as an escape for child marriage, female

genital mutilation, and domestic violence. From an empirical historical lens, and in line with the

results of our paper, this constitutes another suggestive illustration of how encouraging women

to be independent in a safe community might be conducive to a widespread culture of gender

equality.

B Additional Results

This section reports additional estimates corroborating our main findings, relying on both the

OLS and the instrumented specification. Furthermore, we provide results based on propensity

score matching.

OLS Results. As mentioned in the main section, we proxy gender equality using alternative

variables measuring gender differences in literacy levels. Panels A) and B) of Table 7 follow

the same structure of Table 2 with female literacy share and female literacy index measuring

gender equality, respectively. Considered together, these results indicate a strong correlation

between the presence of beguinages and gender equality.

34

Table 7: Beguinages and literacy equality

Baseline Fixed-effects Geography All

(1) (2) (3) (4) (5) (6) (7) (8)

Panel A: Dep. variable: Female lit. share, 1866

Beguinage (0/1) 0.041 0.043 0.036 0.012(0.004)∗∗∗ (0.005)∗∗∗ (0.005)∗∗∗ (0.004)∗∗∗

[0.004]∗∗∗ [0.004]∗∗∗ [0.004]∗∗∗ [0.003]∗∗∗

Exposure (centuries) 0.006 0.006 0.005 0.002(0.001)∗∗∗ (0.001)∗∗∗ (0.001)∗∗∗ (0.001)∗∗

[0.001]∗∗∗ [0.001]∗∗∗ [0.001]∗∗∗ [0.001]∗∗

R2 0.025 0.024 0.204 0.201 0.218 0.214 0.410 0.410

Panel B: Dep. variable: Female lit. index, 1866

Beguinage (0/1) 0.055 0.061 0.048 0.041(0.011)∗∗∗ (0.012)∗∗∗ (0.012)∗∗∗ (0.012)∗∗∗

[0.011]∗∗∗ [0.011]∗∗∗ [0.011]∗∗∗ [0.011]∗∗∗

Exposure (centuries) 0.008 0.007 0.005 0.006(0.002)∗∗∗ (0.002)∗∗∗ (0.002)∗∗ (0.003)∗∗

[0.002]∗∗∗ [0.002]∗∗∗ [0.002]∗∗ [0.002]∗∗

R2 0.005 0.005 0.205 0.203 0.217 0.215 0.234 0.233

Fixed-effects No No Canton Canton Canton Canton Canton CantonGeography No No No No Yes Yes Yes YesDemography No No No No No No Yes YesObservations 2549 2549 2549 2549 2549 2549 2549 2549

Note: This table presents the results of OLS regressions relating the presence of beguinages to increased gender equality proxied by the variable female literacy sharei andfemale literacy index; see main text for a description. Columns 1) and 2) include only the presence of beguinages as a regressor. Columns 3) and 4) add fixed-effects at thecanton level. In Columns 5) and 6) exogenous variables are incorporated, namely, the presence of monasteries, latitude, longitude, distance to Leuven, to the sea and to theclosest river, caloric yield, soil types and crop productivity. Finally, Columns 7) and 8) incorporate demographic variables measured in 1866, including the number of menand women, nuptiality, sex ratio, the share of internal and external migrants and a dummy for the 10% most populous municipalities. Standard errors clustered at thecanton level in parentheses and Conley standard errors, with a Bartlett kernel and a cut-off distance of 20 km, in brackets.∗ p < 0.1, ∗∗ p < 0.05, ∗∗∗ p < 0.01.

35

In Table 8 we introduce a new independent variable, intensityi that combines features of

beguinagei and exposurei. This variable consists of a five-level index formally defined as follows:

Intensityi =

0 if no beguinage has even been recorded,

1 if 1 beguinage has been present and exposurei < 200,

2 if 1 beguinage has been present and exposurei ≥ 200,

3 if more than 1 beguinage has been present and exposurei ≥ 200,

4 if more than 3 beguinages have been present and exposurei ≥ 200.

(10)

The results suggest a positive association between the presence of beguinages and gender equality,

in line with our previous estimates. Moreover, the effect of beguine communities is larger for

intermediate levels of beguinage presence. We also introduce a modified version of the variable

exposurei. This alternative specification considers the total number of years during which any

beguine community was present it a municipality, but does not aggregate the values for those

overlapping in time. Taken together, using alternative outcomes and independent variables gives

more credence to the association between female-only communities and gender equality.

Table 8: Alternative independent variable

Dep. variable: Lit. equality index, 1866

(1) (2)

IntensityNo Beg. Ref1 Beg., < 200 years 0.075

(0.023)∗∗∗

[0.014]∗∗∗

1 Beg., > 200 years 0.151(0.031)∗∗∗

[0.018]∗∗∗

> 1 Beg, > 200 years 0.223(0.057)∗∗∗

[0.037]∗∗

> 3 Beg., > 200 years 0.093(0.024)∗∗∗

[0.037]Alt. exposure (centuries) 0.008

(0.003)∗∗∗

[0.003]∗∗∗

Fixed-effects Canton Canton

Continued on next page

36

Table 8 – Continued from previous page

Dep. variable: Lit. equality index, 1866

(1) (2)

Geography Yes YesDemography Yes YesObservations 2549 2549R2 0.222 0.985

Note: This table presents the results of OLS regressions relating the presence of beguinages to increased genderequality proxied by the variable literacy equality indexi, see main text for a description. The main regressoris the variable intensityi in Column 1), while Column 2) uses a modified version of exposurei. Geography:latitude, longitude, distance to Leuven, to the sea and to rivers, caloric yield, soil type and crop productivity.Demography: population, nuptiality and %age of domestic and international migrants by gender, sex ratioand a dummy for the 10% most populous towns. Regressions include canton fixed-effects. Standard errorsclustered at the canton level in parenthesis and Conley standard errors, with a Bartlett kernel and a cut-offdistance of 20 km in brackets. ∗ p < 0.1, ∗∗ p < 0.05, ∗∗∗ p < 0.01.

Table 9 removes towns with a beguinage from the sample, following Bratti et al. (2017).

Similarly, Table 10 follows the same logic as Table 5 but assumes disconnected networks along

linguistic borders. In general, the results we obtain under these alternative specifications are

qualitatively equivalent to our baseline results, indicating a more gender equal society among

municipalities more connected or exposed to beguinages.

Table 9: Literacy and gender equality, distance: no beguinages

Dep. variable: Lit. equality index, 1866

(1) (2) (3)

Min. Distance Beg. (log-km) −0.010(0.008)[0.007]

IDW 0.087(0.052)∗

[0.044]∗∗

IDW, exposure 0.145(0.046)∗∗∗

[0.039]∗∗∗

Fixed-effects Canton Canton CantonGeography Yes Yes YesDemography Yes Yes YesObservations 2479 2406 2406R2 0.407 0.412 0.414

Continued on next page

37

Table 9 – Continued from previous page

(1) (2) (3)

Note: This table presents the results of OLS regressions relating the presence of beguinages to increasedgender equality proxied by the variable literacy equality indexi, see main text for a description. Column 1)employs distance to the closest beguinage in log-km, Column to 2) introduces IDW as regressor while Column3) uses the IDW weighted using exposure. The sample only comprises towns without a beguinage. Geography:latitude, longitude, distance to Leuven, to the sea and to rivers, caloric yield, soil type and crop productivity.Demography: population, nuptiality and %age of domestic and international migrants by gender, sex ratioand a dummy for the 10% most populous towns. Regressions include canton fixed-effects. Standard errorsclustered at the canton level in parenthesis and Conley standard errors, with a Bartlett kernel and a cut-offdistance of 20 km in brackets. ∗ p < 0.1, ∗∗ p < 0.05, ∗∗∗ p < 0.01.

Table 10: Literacy and gender equality; distance: linguistic borders.

Dep. variable: Lit. equality index, 1866

(1) (2) (3) (4)

IDW, Border 0.067(0.018)∗∗∗

[0.017]∗∗∗

IDW, Border, No Beg. 0.066(0.031)∗∗

[0.028]∗∗

IDW, Border, exposure 0.077(0.021)∗∗∗

[0.018]∗∗∗

IDW, Border, exposure, No Beg. 0.096(0.030)∗∗∗

[0.026]∗∗∗

Fixed-effects Canton Canton Canton CantonGeography Yes Yes Yes YesDemography Yes Yes Yes YesObservations 2474 2405 2474 2405R2 0.438 0.412 0.438 0.414

Note: This table presents the results of OLS regressions relating the presence of beguinages to increasedgender equality proxied by the variable literacy equality indexi, see main text for a description. Column 1)employs IDW as regressor, while Column 3) uses the IDW weighted using exposure. Columns 2) and 4)remove municipalities with a beguinage from the sample and use as regressor those reported in Columns 1)and 3), respectively. In all cases, IDW assumes disconnected networks along linguistic borders. Geography:latitude, longitude, distance to Leuven, to the sea and to rivers, caloric yield, soil type and crop productivity.Demography: population, nuptiality and %age of domestic and international migrants by gender, sex ratioand a dummy for the 10% most populous towns. Regressions include canton fixed-effects. Standard errorsclustered at the canton level in parenthesis and Conley standard errors, with a Bartlett kernel and a cut-offdistance of 20 km in brackets. ∗ p < 0.1, ∗∗ p < 0.05, ∗∗∗ p < 0.01.

In addition, since few municipalities had an operating beguinage in 1886, Panel A) of Table

11 disentangles its effects from the cultural legacy lingering from dismantled communities. We

do so by excluding from the sample municipalities that had an operating beguinage in 1866.

38

This removes 17 municipalities from the baseline sample.61 Results indicate that the cultural

imprint left by beguine communities shaped the perception of women. In general, these results

are indistinguishable from the main specification. Panel B) introduces male literacy rate as

regressor, evidencing that our findings are not caused by women catching up in literacy.62 Male

literacy is always significantly positive. However, the historical presence of beguinages continues

to appear as an important predictor of female agency, although its magnitude decreases and

becomes non-significant for the variable exposurei. Finally, in Panel C), municipal charters,

the instrument used in IV regressions, are included as an additional covariate. Municipal

charters granted towns the possibility of establishing a market and guilds and conveyed other

benefits. Moreover, they indicated relatively prosperous towns. Results evidence that beguine

communities boosted gender equality beyond any possible effect introduced by changes in the

local administration.

Table 11: Beguinages and literacy equality; robustness 2.

Lit. eq. index, 1866

(1) (2)

Panel A: No open beguinage

Beguinage (0/1) 0.047(0.014)∗∗∗

[0.012]∗∗∗

Exposure (centuries) 0.007(0.003)∗∗

[0.003]∗∗

Observations 2539 2539R2 0.429 0.428

Panel B: Male literacy rate

Beguinage (0/1) 0.031(0.012)∗∗∗

[0.011]∗∗∗

Exposure (centuries) 0.004(0.003)[0.002]

Male lit. rate, 1866 0.267 0.268(0.027)∗∗∗ (0.027)∗∗∗

[0.025]∗∗∗ [0.025]∗∗∗

Continued on next page

61The municipalities that had an ongoing beguinage in 1866 are, with the closing year between brackets: Aalst(1953), Kortijk (2001), Diksmuide (1914), Hasselt (1886), Herenthals (2001), Momalle (1899), Hoogstraten (1972),Liege (1869), Lier (1970), Nieuwpoort (1914) and Turnhout (1994).

62Notice that male literacy rates may be an outcome variable of historical beguinal presence.

39

Table 11 – Continued from previous page

Lit. eq. index, 1866

(1) (2)

Observations 2549 2549R2 0.469 0.469

Panel C: Municipal charter

Beguinage (0/1) 0.038(0.016)∗∗

[0.014]∗∗∗

Exposure (centuries) 0.004(0.003)[0.003]

Municipal charter 0.018 0.027(0.020) (0.018)[0.020] [0.018]

Observations 2549 2549R2 0.433 0.432

Fixed-effects Canton CantonGeography Yes YesDemography Yes Yes

Note: This table presents the results of OLS regressions relating the presence of beguinages to increased genderequality. Panel A) excludes ongoing beguinages. Panel B) and C) control for male literacy and municipalcharters, respectively. Geography: latitude, longitude, distance to Leuven, to the sea and to rivers, caloric yield,soil types and crop productivity. Demography: population, nuptiality and %age of domestic and internationalmigrants by gender, sex ratio and a dummy for the 10% most populous towns. Regressions include cantonfixed-effects. Standard errors clustered at the canton level in parenthesis and Conley standard errors, with aBartlett kernel and a cut-off distance of 20 km in brackets. ∗ p < 0.1, ∗∗ p < 0.05, ∗∗∗ p < 0.01.

We also follow Valencia Caicedo (2018) and compare the effect of beguinages that were

abandoned soon after they were established to those that remained. Table 12 presents the results

of regressions including separate indicators for beguinages within the first, 10th, 20th and 30th

centiles of the distribution of exposure (in Column 1), 2), 3) and 4), respectively) and for longer

durations. This specification illustrates that our results are not driven by specific gender-related

attitudes that were specific to towns with a beguinage ex-ante. In other words, assuming that

beguines selected their location based on the local residents’ attitudes towards women, one

would expect locations where beguinages were short lived and those where beguine communities

endured longer to display very similar gender norms. Hence, if our previous estimates captured

only a pre-existing gender equality differential across municipalities, also shortly lived beguinages

should appear as predictors of gender outcomes in 1886. The lack of such results reinforces the

interpretation that beguines did trigger a change in gender equality that greatly persisted over

time.

40

Table 12: Municipal charters and gender equality.

Dep. var.: Literacy Equality Index, 1866

1st cent. 10th cent. 20th cent. 30th cent.(1) (2) (3) (4)

Short beg. 0.025 0.032 0.035 0.037(0.032) (0.028) (0.018)∗ (0.014)∗∗

[0.031] [0.027] [0.017]∗∗ [0.014]∗∗∗

Long. beg. 0.045 0.044 0.045 0.046(0.014)∗∗∗ (0.014)∗∗∗ (0.016)∗∗∗ (0.018)∗∗∗

[0.012]∗∗∗ [0.013]∗∗∗ [0.014]∗∗∗ [0.016]∗∗∗

Observations 2549 2549 2549 2549R2 0.429 0.429 0.429 0.429

Fixed-effects Yes Yes Yes YesGeography Yes Yes Yes YesDemography Yes Yes Yes YesNr of Short beg. 6 7 14 21

Note: This table presents the results of OLS regressions relating the presence of beguinages to increased genderequality. Gender equality is proxied by literacy equality index. Column 1) includes separate indicators forbeguinages with exposure in the first centile of the distribution. Columns 2), 3) and 4) increase it to the 10th,20th and 30th centiles, respectively. Geography: latitude, longitude, distance to Leuven, to the sea and torivers, caloric yield, soil types and crop productivity. Demography: number of men and women, marriagerate, the sex ratio, the percentage of domestic and international migrants by gender and an indicator variabletaking value one for municipalities above the 10th percentile in terms of population. Arrondissement andcanton fixed-effects in Column 1) and 2), respectively since it was not possible to estimate a model in Column1) with canton fixed-effects. Standard errors clustered at the canton level. ∗ p < 0.1, ∗∗ p < 0.05, ∗∗∗ p < 0.01.

IV Results. First, in Table 13, we present some evidence regarding the nil effect of municipal

charters on gender equality among towns that had arguably similar attitudes toward women.

We do so by analysing the effect of municipal charters on two sub-samples: towns that hosted a

beguinge, and towns that never had one. Arguably, if beguines self-selected towns because they

were more tolerant towards their lifestyle, attitudes toward women should be similar within each

sub-sample. Therefore, we ascertain the effect of municipal charters on gender equality while

holding attitudes constant. According to our results, municipal charters did not boost gender

equality, giving additional credence to our identification strategy. This result is similar to Panel

C) of Table 11.

We also present the results for the alternative robustness checks performed on the IV

specification. In particular, Table 14 restricts the sample to municipalities closely located around

towns with a beguinage at distances 5 km, 10 km and 20 km, while Table 15 replicates Panel A)

and B) of Table 11. In Panel A) of Table 16 the main regressor is the logarithm of the distance

from each municipality to the closest beguinage, instrumented by municipal charters. Panel B)

41

Table 13: Municipal charters and gender equality.

Dep. variable: Fem. Lit. Eq.

Munic. with beguinage Munic. without beguinage(1) (2)

Municipal charter −0.008 0.034(0.078) (0.035)[0.024] [0.032]

Fixed-effects Arrond. CantonGeography Yes YesDemography Yes YesObservations 70 2479R2 0.978 0.404

Note: This table relates municipal charters with gender equality using OLS regressions. Gender equality isproxied by the variable literacy equality indexi. See the main text for a description. Column 1) includes allmunicipalities that hosted a beguine community at any moment in time while Column 2) considers thosethat never hosted one. Geography: latitude, longitude, distance to Leuven, to the sea and to rivers, caloricyield, soil types and crop productivity. Demography: number of men and women, marriage rate, the sex ratio,the percentage of domestic and international migrants by gender and an indicator variable taking value onefor municipalities above the 10th percentile in terms of population. Arrondissement and canton fixed-effectsin Column 1) and 2), respectively since it was not possible to estimate a model in Column 1) with cantonfixed-effects. Standard errors clustered at the canton level. ∗ p < 0.1, ∗∗ p < 0.05, ∗∗∗ p < 0.01.

introduces the alternative definition for exposurei,j as the main regressor.

Table 14: Beguinages and Literacy equality; IV, robustness 1.

Dep variable: Lit. eq. index, 1866

5Km 10Km 20Km

(1) (2) (3) (4) (5) (6)

Beguinage (0/1) 0.053∗∗∗ 0.090∗∗∗ 0.097∗∗∗

(0.018) (0.025) (0.024)Exposure (centuries) 0.009∗ 0.008∗ 0.013∗∗

(0.005) (0.004) (0.006)

Observations 301 371 723 1114 955 2060R2 0.633 0.659 0.516 0.516 0.464 0.4341st-stage F-val. 213.1 24.4 147.4 3.7 . 153.7 3.8

Fixed-effects Canton Canton Canton Canton Canton CantonGeography Yes Yes Yes Yes Yes YesDemography Yes Yes Yes Yes Yes Yes

Continued on next page

42

Table 14 – Continued from previous page

Dep variable: Lit. eq. index, 1866

5Km 10Km 20Km

(1) (2) (3) (4) (5) (6)

Note: This table presents the results of IV regressions where the presence of beguine communities is instrumentedusing municipal charters. Columns 1) and 2) consider only municipalities located less than 5 km from abeguinage, Columns 3) and 4) increase the distance up to 10 km and Columns 5) and 6) further expand itto 20 km. Geography: latitude, longitude, distance to Leuven, to the sea and to rivers, caloric yield, soiltypes and crop productivity. Demography: number of men and women, nuptiality, sex ratio, the percentage ofdomestic and international migrants and a dummy for the 10% most populous municipalities. Standard errorsclustered at the canton level. ∗ p < 0.1, ∗∗ p < 0.05, ∗∗∗ p < 0.01.

Table 15: Beguinages and literacy equality; IV, robustness 2.

Lit. eq. index, 1866 Fem. lit. share, 1866 Fem. lit. index, 1866

(1) (2) (3) (4) (5) (6)

Panel A: No open beguinage

Beguinage (0/1) 0.096∗∗∗ 0.025∗∗∗ 0.089∗∗∗

(0.021) (0.006) (0.021)Exposure (centuries) 0.019∗∗ 0.006∗∗ 0.019∗∗

(0.008) (0.002) (0.008)

Observations 831 2539 831 2539 831 2539R2 0.463 0.424 0.440 0.403 0.245 0.2261st-stage F-val. 163.0 14.3 163.0 14.3 163.0 14.3

Panel B: Male literacy rate

Beguinage (0/1) 0.057∗∗∗ 0.012∗∗ 0.050∗∗∗

(0.017) (0.005) (0.017)Exposure (centuries) 0.008∗ 0.002 0.008

(0.005) (0.001) (0.005)Male lit. rate, 1866 0.205∗∗∗ 0.268∗∗∗ 0.069∗∗∗ 0.093∗∗∗ 0.213∗∗∗ 0.280∗∗∗

(0.035) (0.026) (0.011) (0.009) (0.037) (0.027)

Observations 973 2549 973 2549 973 2549R2 0.494 0.468 0.475 0.455 0.296 0.2831st-stage F-val. 197.4 30.5 197.4 30.5 197.4 30.5

Fixed-effects Canton Canton Canton Canton Canton CantonGeography Yes Yes Yes Yes Yes YesDemography Yes Yes Yes Yes Yes Yes

Note: This table presents the results of IV regressions relating the presence of beguinages to increased genderequality where the presence of beguine communities is instrumented using municipal charters. Panel A)removes ongoing beguinages from the sample. Panel B) includes male literacy rates as additional control.Geography: latitude, longitude, distance to Leuven, to the sea and to rivers, caloric yield, soil types and cropproductivity. Demography: number of men and women, nuptiality, sex ratio, the percentage of domestic andinternational migrants and a dummy for the 10% most populous municipalities. Standard errors clustered atthe canton level. ∗ p < 0.1, ∗∗ p < 0.05, ∗∗∗ p < 0.01.

43

Table 16: Beguinages and literacy equality; IV, robustness 3.

Lit. eq. index, 1866 Fem. lit. share, 1866 Fem. lit. index, 1866

(1) (2) (3)

Panel A: Distance to beguinage as regressor

Dist. to closest beg. −0.038∗∗ −0.011∗∗ −0.038∗∗

(log-km) (0.016) (0.005) (0.016)

Observations 2549 2549 2549R2 0.426 0.405 0.2261st-stage F-val. 34.7 34.7 34.7

Panel B: Alternative definition of exposure

Exposure (centuries) 0.013∗∗ 0.004∗∗ 0.013∗∗

(0.005) (0.002) (0.005)

Observations 2549 2549 2549R2 0.431 0.409 0.2321st-stage F-val. 35.547 35.547 35.547

Fixed-effects Canton Canton CantonGeography Yes Yes YesDemography Yes Yes Yes

Note: This table presents the results of IV regressions relating the presence of beguinages to increased genderequality; the presence of beguine communities is instrumented using municipal charters. Panel A) featuresas the main independent variable the distance to the closest beguinal community, in log-km. See main textfor a discussion regarding the benefits of a continuous variable. Panel B) employs a modified version of thevariable exposurei. See main text for a description. Geography: latitude, longitude, distance to Leuven, tothe sea and to rivers, caloric yield, soil types and crop productivity. Demography: number of men and women,nuptiality, sex ratio, the share of internal and external migrants and a dummy for the 10% most populousmunicipalities. Standard errors clustered at the canton level. ∗ p < 0.1, ∗∗ p < 0.05, ∗∗∗ p < 0.01.

Propensity score matching. In order to identify the effect of the presence of beguine

communities on gender-related outcomes while reducing the bias due to confounding variables,

we implement propensity score matching on our original sample. Ideally, treated and control

municipalities should be matched using some pre-beguinage characteristics on attitudes towards

women at the municipal level. Unfortunately, this information is not available. Instead,

municipalities are first matched on exogenous covariates thus mitigating concerns on the threat

to identification posed by unobservables. These are latitude and longitude, distance to the

sea and to rivers and cereal productivity. In a second matching exercise, predictors include

the distance to Leuven and the presence of monasteries at the municipal level to consider the

regional organisation of power and wealth concentration.

Table 17 displays the positive relationship between the historical presence of beguine com-

munities and gender equality obtained with this alternative estimation approach. It establishes

44

Table 17: Propensity score matching.

Lit. eq. index, 1866

Basic Matching Extended Matching(1) (2)

Stratified matching 0.078∗∗ 0.052∗

(0.031) (0.027)

Observations 2560 2560Treated 70 70Controls 1542 1513

Kernel matching 0.080∗∗∗ 0.063∗∗

(0.028) (0.027)

Observations 2560 2560Treated 70 70Controls 2483 2483

Nearest neig. matching 0.097∗∗∗ 0.071∗∗

(0.025) (0.032)

Observations 2560 2560Treated 70 70Controls 44 45

Note: This table presents the results of propensity score matching relating the presence of beguinages toincreased gender equality. Matches are identified using stratified, kernel and nearest neighbour matching,respectively. Matches are performed using latitude, longitude, distance to the closest river and to the sea, andcrop-specific productivity in Column 1). Column 2) additionally employs the presence of monasteries and thedistance to Leuven to match municipalities.∗ p < 0.1, ∗∗ p < 0.05, ∗∗∗ p < 0.01.

that beguine presence at the local level left a great imprint on gender equality, still measurable

in the 19th century. In general, the magnitude of coefficients is slightly larger the one of the

baseline OLS specification presented in Column 7) of Table 2.

45

ISSN 1379-244X D/2019/3082/13